Friday, July 16, 2021

Mythbuster Video #6 -- On to 20 Meters (But With Bandpass Filter Woes). Please help solve the mystery!

Here's how I started with the Elsie program. Note that to get a 50 ohm match on both ends it needs an impractically low value for the coils (.064 uH). 

But Elsie lets you specify the coil value.  So I then I went with 1 uH.  But with this value you don't get 50 ohms at either end.  You need a matching network.  Elsie provides this too!  

I asked Elsie to match my BP filter to 50 ohms.  It provided several options to do this --  I went with a simple capacitive impedance divider.  But alas, I was now bumping up against the 7 limit of the free version of Elsie so I had to reduce the number of LC elements from 4 to 3.  Bummer. 

With 3 LC tuned circuits and matched to 50 ohms the plot looks OK.  But I would have preferred 4 LC circuits. 
 

The rftools website created a BP filter for me with 4 LC elements, and matched to 50 ohms.  Very useful. https://rf-tools.com/lc-filter/

But here's my problem:  With both the filter designed by Elsie and the one designed by rftools, I found that the filter passband was too low.  It was in the 12 - 13 MHz range.   I found that by removing 3 turns from the 1 uH coils I could move the passband up to the desired range.  But why the discrepancy?  I was measuring the coils and the caps with an AADE meter.  I was testing the passband both with a NanoVNA and with a combination of an HP8640B sig gen and a Rigol oscilloscope (with the filter terminated into a 50 ohm resistor).  Any suggestions on why these filters should have passbands lower than predicted would be appreciated. 

10 comments:

  1. Hi Bill, great project you are building.
    One of the reasons your passband was to low is that you might have created some stray capacity between your coils and the PCB, 73 de Peter VA6BHP

    ReplyDelete
    Replies
    1. I don't think so. I measured it today. Just the coil: 1.111 uH. Coil connected between a Manhattan pad and ground, vertical to the board: 1.169 uH . Same coil horizontal, lying flat on the copper: 1.174 uH. By compressing the turns a bit I can increase the inductance to 1.214 uH. However, the pads themselves do seem to add significant capacitance to ground -- 8.74 pF. I'll have to try this with different pads. 73 Bill

      Delete
  2. I haven't been following the build but is it possible that the stages on either of the BPF aren't 50 ohms?

    ReplyDelete
    Replies
    1. I don't think that's it. I was measuring into a 50 ohm resistor. 73 Bill

      Delete
    2. Ah you said that in the article. One day I will learn to read. 73

      Delete
  3. Rx sounds sweet! If you have resonant antennas, no RF gain is necessary below 20m. I do use a moderate gain RF preamp, 8 to 12dB, on 20m. With only two TIA amps in the IF I would expect the Rx to be slightly under powered above 10mhz. If you add more IF gain, you start to really need AGC!

    Ive also found Elsie suggests unusable L and C values. It can be frustrating and on a few occasions I've reverted to proven BPFs.

    No idea why your filters are low. One thing though, my inductors made according to toroids.info are usually 10% low, which is ok because it's easy to prune them using an LC meter. I don't think stray or pad capacitance would be significant at 40m, may be 20m, have you measured the C exhibited by a pad with your AADE meter? -- VK3HN.

    ReplyDelete
    Replies
    1. Paul: The pads themselves do seem to add significant capacitance to ground -- 8.74 pF. I'll have to try this with different pads. 73 Bill

      Delete
  4. This comment has been removed by the author.

    ReplyDelete
  5. The radio sounds great!

    I entered you 4 pole filter into LTSPICE, and as you indicated, it simulated fine. You mentioned that you were seeing about 9 pF from your pads. so, I added 9pF to all of the caps that tie directly to GND to simulate this added capacitance, and the filter was shifted by down quite a bit just as you reported, so I think it is you construction technique adding the capacitance.

    If you build your filters on single sided PCB material and cut round pads in the foil to solder to you can avoid this capacitance. You can modify a woodworking forstner bit to do this for you, I've also seen bits that are sold from time to time. Also, you'll find that a lot of printed circuit boards for filters remove the ground fill around filter components to avoid stray capacitance. Or, add variable caps (or inductors) into the design to tune the filters to the final band pass characteristic. This is much easier than it used to be with the advent of the NanoVNA's!

    ReplyDelete
  6. Thanks to all who provided suggestions on the BP filters. I think I was too quick to dismiss the idea that the copper on the board was changing the inductance -- I now think it was. With the coils placed on the board I had to remove 3 turns to get them to 1 uH. I eliminated the pad capacitance problem by going "dead bug" in the filter construction. Four LC filters may be a bit some problematic overkill. I tried to make sure that each of them resonated separately at the freq indicated by the L and C values. This helped, but when I put the four of them together, the filter passband was too high. So I squeezed the turns on the toroids together while watching on the NanoVNA -- doing this I was able to get the desired passband. Next time: No more than three LC circuits, and trimmer caps to ease in the inevitable tuning. 73 Bill

    ReplyDelete